近年来,随着监控摄像头的不断增多和互联网的迅速发展,监控视频与网络视频越来越多,对视频进行自动行为冲突检测对降低人为审核导致的隐私信息泄露风险及维护社会治安、净化网络环境等具有重要意义.为了充分提取视频中的行为冲突特征,...近年来,随着监控摄像头的不断增多和互联网的迅速发展,监控视频与网络视频越来越多,对视频进行自动行为冲突检测对降低人为审核导致的隐私信息泄露风险及维护社会治安、净化网络环境等具有重要意义.为了充分提取视频中的行为冲突特征,并获得有较好泛化能力与检测效果的模型,采用I3D(inflated 3D convolutional network)与VGGish,基于XD-Violence进行多模态特征的提取,并提出了基于Transformer和图卷积网络的行为冲突检测模型TG-BCDM(behavior conflict detection model based on Transformer and graph convolution networks).该模型包含Transformer编码器模块和图卷积模块,可以在有效捕捉视频中长距离依赖关系的同时,关注视频特征的全局信息和局部信息.经过实验证明,该模型优于现有的8种方法.展开更多
This study investigated the influence factors on the seismic response and deformation modes of retaining walls using large-scale model shaking table tests. Experimental results showed that the distribution of peak sei...This study investigated the influence factors on the seismic response and deformation modes of retaining walls using large-scale model shaking table tests. Experimental results showed that the distribution of peak seismic earth pressures along the height of a wall was a single peak value curve. The seismic earth pressures on a gravel soil retaining wall were larger than the pressures on the weathered granite and quartz retaining walls. Also, the peak seismic earth pressure increased with increases in the peak ground acceleration and the wall height. The measured seismic active earth pressures on a rock foundation retaining wall were larger than the calculated values, and the action position of resultant seismic pressure was higher than 0.33 H. In the soil foundation retaining wall, the measured seismic earth pressures were much smaller than the calculated values, while the action position was slightly higher than 0.33 H. The soil foundation retaining wall suffered base sliding and overturning under earthquake conditions, while overturning was the main failure mode for the rock foundation retaining walls.展开更多
文摘近年来,随着监控摄像头的不断增多和互联网的迅速发展,监控视频与网络视频越来越多,对视频进行自动行为冲突检测对降低人为审核导致的隐私信息泄露风险及维护社会治安、净化网络环境等具有重要意义.为了充分提取视频中的行为冲突特征,并获得有较好泛化能力与检测效果的模型,采用I3D(inflated 3D convolutional network)与VGGish,基于XD-Violence进行多模态特征的提取,并提出了基于Transformer和图卷积网络的行为冲突检测模型TG-BCDM(behavior conflict detection model based on Transformer and graph convolution networks).该模型包含Transformer编码器模块和图卷积模块,可以在有效捕捉视频中长距离依赖关系的同时,关注视频特征的全局信息和局部信息.经过实验证明,该模型优于现有的8种方法.
基金the National Program on Key Research Project of China (Grant No. 2016YFC0802206)the open research fund of MOE Key Laboratory of High-speed Railway Engineering,Southwest Jiaotong University and Doctoral Innovation Fund Program of Southwest University of Science and Technology (Grant No. 16zx7123)
文摘This study investigated the influence factors on the seismic response and deformation modes of retaining walls using large-scale model shaking table tests. Experimental results showed that the distribution of peak seismic earth pressures along the height of a wall was a single peak value curve. The seismic earth pressures on a gravel soil retaining wall were larger than the pressures on the weathered granite and quartz retaining walls. Also, the peak seismic earth pressure increased with increases in the peak ground acceleration and the wall height. The measured seismic active earth pressures on a rock foundation retaining wall were larger than the calculated values, and the action position of resultant seismic pressure was higher than 0.33 H. In the soil foundation retaining wall, the measured seismic earth pressures were much smaller than the calculated values, while the action position was slightly higher than 0.33 H. The soil foundation retaining wall suffered base sliding and overturning under earthquake conditions, while overturning was the main failure mode for the rock foundation retaining walls.
